Skip to content

README.md

Latest commit

 

History

History
225 lines (166 loc) · 4.36 KB

README.md

File metadata and controls

225 lines (166 loc) · 4.36 KB

Examples

This directory contains practical examples demonstrating Context usage patterns.

Examples Overview

1. Backend Automation

Path: backend-automation/

Demonstrates using Context in backend automation scenarios like GitHub Actions or scheduled jobs.

Features:

  • Creating contexts for data analysis
  • Adding inputs with relevance scores
  • Automatic pruning to fit token limits
  • Strategy-based routing
  • Context operations (extend, merge, serialize)

Run:

cd backend-automation
PYTHONPATH=../../core/python python backend_example.py

Use Cases:

  • Scheduled catalog analysis
  • Automated data processing
  • Batch LLM operations
  • Report generation

2. Browser Analysis

Path: browser-analysis/

Demonstrates using Context in browser environments with user-provided API keys.

Features:

  • Browser-based Context creation
  • Interactive input management
  • Strategy selection
  • Context serialization
  • Stub execution (shows structure without actual API calls)

Run:

cd browser-analysis
# Open index.html in a browser
python -m http.server 8000
# Then visit http://localhost:8000

Use Cases:

  • User-driven data exploration
  • Client-side analysis with user keys
  • Interactive data visualization
  • Privacy-preserving analysis

3. Hybrid Workflow

Path: hybrid-workflow/

Demonstrates cross-runtime workflow: backend creates context, frontend extends and customizes.

Features:

  • Backend preprocessing and initial analysis
  • Context serialization for frontend
  • Frontend context loading and extension
  • User customization with their API keys
  • Context sharing via URL or download

Run Backend:

cd hybrid-workflow
PYTHONPATH=../../core/python python backend.py

Run Frontend:

# Open frontend.html in a browser
python -m http.server 8000
# Then visit http://localhost:8000/frontend.html

Use Cases:

  • Collaborative analysis workflows
  • Backend preprocessing + frontend customization
  • Shareable analysis contexts
  • Reproducible research

Common Patterns

Creating a Context

from context import Context

ctx = Context(
    intent="analyze",
    constraints={"max_tokens": 4000},
    routing={"strategy": "cost_optimized"}
)

Adding Inputs

# Add with relevance score
ctx.add_input(data, relevance=0.9)

# Multiple inputs
for item in dataset:
    ctx.add_input(item, relevance=item.priority)

Pruning

# Prune to fit token limit
ctx.prune(max_tokens=2000, relevance_threshold=0.5)

Routing

# Strategy-based
ctx.route(strategy="cost_optimized")

# Explicit
ctx.route(model="gpt-4", provider="openai")

Execution

result = ctx.execute(
    task="Analyze this data",
    api_key=os.environ["OPENAI_API_KEY"]
)

Serialization

# Save
json_str = ctx.to_json()
with open("context.json", "w") as f:
    f.write(json_str)

# Load
with open("context.json") as f:
    ctx = Context.from_json(f.read())

Cross-Runtime

Backend (Python):

ctx = Context(intent="analyze")
ctx.add_input(preprocessed_data)
# Save for frontend
with open("context.json", "w") as f:
    f.write(ctx.to_json())

Frontend (TypeScript):

// Load context created by backend
const response = await fetch('context.json');
const data = await response.json();
const ctx = Context.fromJSON(data);

// Extend with user data
ctx.addInput(userInput);
const result = await ctx.execute({
    task: "Custom analysis",
    apiKey: userApiKey
});

Running Examples

All examples use stub execution by default (no real API calls). To enable real execution:

  1. Set up API keys:

    export OPENAI_API_KEY="your-key-here"
# or
export ANTHROPIC_API_KEY="your-key-here"

  2. Implement provider adapters in adapters/ directory

  3. Update executor to use real providers instead of stubs

Example Data

Examples use synthetic data for demonstration. In real usage, replace with your actual data sources:

  • Catalog metadata
  • User inputs
  • API responses
  • Database queries
  • File contents

Next Steps

After reviewing examples:

  1. Try modifying the examples for your use case
  2. Review the Getting Started Guide
  3. Check the API Reference
  4. Read the Migration Guide for adopting Context in existing projects